Yangzihao Wang, Ph.D.

Wang Yangzihao is a founding member and research engineer at Sea AI Lab working on AI system and neural graphics. From 2018 to 2021 he worked as a senior software engineer at WeChat on infrastructure for high-performance graph computing, graph representation learning systems and large-scale recommender system and worked as a senior research engineer at Tencent on AI plaform and AutoML for reinforcement learning. From 2017 to 2018 he worked as a software engineer at Google Brain’s TensorFlow team on distributed training and performance engineering. He has published on conferences such as NeurIPS, SIGIR, PPoPP, and IPDPS.

Dr. Wang graduated with a Ph.D. in computer science from UC Davis in 2016. During his PhD years, he worked with Prof. John Owens on various research topics including: 1) structure of parallelism and locality in irregular algorithms such as graph algorithms on the GPU; 2) parallel programming model for graph analytics; and 3) large-scale graph processing and data analysis system. He also did internships at AMD Research, DARPA, and Google.

Before UC Davis, Dr. Wang received B.E. degree in computer science in 2007 and M.E. degree in software engineering in 2011 both from Beijing University of Aeronautics and Astronautics. During his Master's years, he worked on several projects on physical simulation and distributed rendering systems.

Projects[^]

HloEnv:Compiler optimization should be re-defined using deep learning. This is the first step towards this goal. An environment based on XLA for deep learning compiler optimization research. It's on github too.

Plato:A framework for distributed graph computation and machine learning at WeChat scale. It's on github. Stay tuned! More graph computing projects at WeChat/Tencent to be open-sourced!

Gunrock:Fastest graph processing library on the GPU. The library is on github.

CUDA Depixelizer: A de-pixelization algorithm on GPU to vectorize a pixel art into a vector graph based on this paper. Source code can be found on github.

Zhua: A forum crawler using Python Scrapy. My first practical Python project so far. Made with love for my lovely wife on her research topic which focuses on using data analysis and communication theory to help depressed patients. Source code can be found on github.

Dissertation[^]

Gunrock: A Programming Model and Implementation for Graph Analytics on Graphics Processing Units.

Publications[^]

Peer-Reviewed Papers:

Oh Chin-Yang, Kunhao Zheng, Bingyi Kang, Xinyi Wan, Zhongwen Xu, Shuicheng Yan, Min Lin, Yangzihao Wang HloEnv: A Graph Rewrite Environment for Deep Learning Compiler Optimization Research. Accepted by Workshop on ML for Systems at NeurIPS 2022, Dec 2022. [ pdf ]

Haidong Rong, Yangzihao Wang, Feihu Zhou, Junjie Zhai, Haiyang Wu, Rui Lan, Fan Li, Han Zhang, Yuekui Yang, Zhenyu Guo, Di Wang Distributed Equivalent Substitution Training for Large-Scale Recommender Systems. Accepted by Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'20), Jul 2020. [ DOI | talk ]

Xianyan Jia, Shutao Song, Wei He, Yangzihao Wang, Haidong Rong, Feihu Zhou, Liqiang Xie, Zhenyu Guo, Yuanzhou Yang, Liwei Yu, Tiegang Chen, Guangxiao Hu, Shaohuai Shi, Xiaowen Chu. Highly Scalable Deep Learning Training System with Mixed-Precision: Training ImageNet in Four Minutes. Workshop on Systems for ML and Open Source Software at NeurIPS 2018, December 2018.[ http ]

Yuechao Pan, Yangzihao Wang, Yuduo Wu, Carl Yang, and John D. Owens. Multi-GPU Graph Anlytics. In Proceedings of the 31st IEEE International Parallel and Distributed Processing Symposium, IPDPS 2017, May/June 2017. [ bib | http ]

Yangzihao Wang, Sean Baxter, John D. Owens. Mini-Gunrock: A Lightweight Graph Analytics Framework on the GPU. In Graph Algorithms Building Blocks, GABB 2017, May 2017. [ bib | http ]

Leyuan Wang, Yangzihao Wang, Carl Yang, and John D. Owens. A Comparative Study on Exact Triangle Counting on the GPU. In Proceedings of the 1st High Performance Graph Processing Workshop, HPGP'16, May 2016. [ bib | DOI | http ]

Yangzihao Wang, Andrew Davidson, Yuechao Pan, Yuduo Wu, Andy Riffel, and John D. Owens. Gunrock: A high-performance graph processing library on the GPU. In Proceedings of the 21th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2016, March 2016, Distinguished Paper Award [ bib | DOI | http ]

Yuduo Wu, Yangzihao Wang, Yuechao Pan, Carl Yang, and John D. Owens. Performance Characterization for High-Level Programming Models for GPU Graph Analytics. In IEEE International Symposium on Workload Characterization, IISWC2015, October 2015 [ bib | http ]

Carl Yang, Yangzihao Wang, and John D. Owens. Fast Sparse Matrix and Sparse Vector Multiplication Algorithm on the GPU. To appear in IPDPS Graph Algorithms Building Blocks Workshop 2015, GABB 2015 [ bib | http ]

Afton Geil, Yangzihao Wang, and John D. Owens. WTF, GPU! computing twitter's who-to-follow on the GPU. In Proceedings of the Second ACM Conference on Online Social Networks, COSN '14, pages 63-68, New York, NY, USA, 2014. ACM. [ bib | DOI | http ]

Non-Peer-Reviewed Papers:

Wanjing Wei, Yangzihao Wang, Pin Gao, Shijie Sun, Haidong Yu A Distributed Multi-GPU System for Large-Scale Node Embedding at TencentCoRR, /abs/2005.13789, Jun 2020. [ arXiv ]

Haidong Rong, Yangzihao Wang, Feihu Zhou, Junjie Zhai, Haiyang Wu, Rui Lan, Fan Li, Han Zhang, Yuekui Yang, Zhenyu Guo, Di Wang Distributed Equivalent Substitution Training for Large-Scale Recommender SystemsCoRR, /abs/1909.04823, Sept 2019. [ arXiv ]

Xianyan Jia, Shutao Song, Wei He, Yangzihao Wang, Haidong Rong, Feihu Zhou, Liqiang Xie, Zhenyu Guo, Yuanzhou Yang, Liwei Yu, Tiegang Chen, Guangxiao Hu, Shaohuai Shi, Xiaowen ChuHighly Scalable Deep Learning Training System with Mixed-Precision: Training ImageNet in Four MinutesCoRR, abs/1807.11205(1807.11205v1), July 2018. [ bib | arXiv ]

Yangzihao Wang, Yuechao Pan, Andrew Davidson, Yuduo Wu, Carl Yang, Leyuan Wang, Muhammad Osama, Chenshan Yuan, Weitang Liu, Andy T. Riffel, and John D. Owens.Gunrock: GPU Graph Analytics.CoRR, abs/1701.01170(1701.01170v1), January 2017. [ bib | arXiv ]

Yuechao Pan, Yangzihao Wang, Yuduo Wu, Carl Yang, and John D. Owens. Multi-GPU Graph Analytics. CoRR, abs/1504.04804(1504.04804v1), April 2015. [ bib | arXiv ]

Yangzihao Wang, Andrew Davidson, Yuechao Pan, Yuduo Wu, Andy Riffel, and John D. Owens. Gunrock: A high-performance graph processing library on the GPU. CoRR, abs/1501.05387(1501.05387v6), March 2015. [ bib | arXiv ]

Yangzihao Wang, Yuduo Wu Scene Classification with Deep Convolutional Neural Networks. Tech Report, Dec 2015. [ http ]

Yangzihao Wang and John D. Owens. Large-Scale Graph Processing Algorithms on the GPU Technical Report, January 2013.[pdf ]

Posters:

Leyuan Wang, Yangzihao Wang, John D. Owens. Fast Parallel Subgraph Matching on the GPU In Proceedings of the 1st High Performance Graph Processing Workshop, HPGP'16, May 2016. [ poster ]

Yangzihao Wang, Andrew Davidson, Yuechao Pan, Yuduo Wu, Andy Riffel, and John D. Owens. Gunrock: A high-performance graph processing library on the GPU. In Proceedings of the 20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2015, pages 265-266, New York, NY, USA, 2015. ACM. [ bib | DOI | poster ]

Yuechao Pan, Yangzihao Wang, Yuduo Wu, Carl Yang, and John D. Owens. Multi-GPU Graph Analytics. Supercomputing 2015. [ poster ]

Yangzihao Wang, Andrew Davidson, Yuechao Pan, Yuduo Wu, Andy Riffel, and John D. Owens. Gunrock: A high-performance graph processing library on the GPU. In Proceedings of the 24th International Conference on Parallel Architectures and Compilation Techniques, PACT 2015. [ poster ]

Talks[^]

Training ImageNet in Four Minutes
GPU Technology Conference Silicon Valley, March 2019 slides | video
Gunrock: A Fast and Programmable Multi-GPU Graph Processing Library
GPU Technology Conference, April 2016 slides | video
High-Performance Graph Primitives on GPUs: Design and Implementation of Gunrock
Oracle Labs, April 2015 slides

NVIDIA Research, Jan 2015 slides

GPU Technology Conference, March 2014 slides | video